ABSTRACT
Parkinsonian motor deficits are associated with elevated inhibitory output from the basal ganglia (BG). However, several features of Parkinson’s disease (PD) have not been accounted for by this supra-inhibition framework, including the potentially therapeutically-relevant observation that movements guided by external stimuli are less impaired than otherwise-identical movements generated based on internal goals. Is this difference due to divergent processing within the BG itself, or to the recruitment of extra-BG pathways by sensory processing? In addition, surprisingly little is known about precisely when, in the sequence from selecting to executing movements, BG output is altered by PD. Here, we address these questions by recording activity in the SNr, a key BG output nucleus, in hemiparkinsonian (hemi-PD) mice performing a well-controlled behavioral task requiring stimulus-guided and internally-specified directional movements. We found that hemi-PD mice (n = 5, male) exhibited a bias ipsilateral to the side of dopaminergic cell loss, consistent with supra-inhibition of contralateral movements by BG output, and that this bias was stronger when movements were internally specified rather than stimulus guided, consistent with clinical observations in parkinsonian patients. We further found that changes in SNr activity during movement preparation could account for the ipsilateral behavioral bias, as well as its greater magnitude for internally-specified movements. These results suggest that parkinsonian changes in BG output underlying movement preparation contribute to the greater deficit in internally-specified in comparison to stimulus-guided movements.
Significance Statement Parkinsonian patients exhibit the intriguing phenomenon that movements guided by external stimuli are often less impaired than otherwise-identical movements generated based on internal goals. For example, patients can exhibit a more normal gait when their steps are guided by patterned floor tiling than when traversing a featureless floor. Whether this difference in movement execution is due to distinct processing intrinsic to the basal ganglia (BG) or to compensation from other motor pathways is an open question with therapeutic implications. We addressed this question by recording BG output during behavior in a parkinsonian mouse model. We found that mice exhibited greater impairment in internally-specified than stimulus-guided movements, and that differences in BG output during movement preparation could account for this effect.
Introduction
Parkinson’s disease (PD) is a neurodegenerative disease of the basal ganglia (BG) in which motor impairments arise from disordered – typically, elevated – inhibitory BG output resulting from the loss of dopaminergic tone (DeLong, 1990; Wichmann et al., 1999; Ibanez-Sandoval et al., 2007; Utter and Basso, 2008; Wang et al., 2010a; Seeger-Armbruster and von Ameln-Mayerhofer, 2013; Brazhnik et al., 2014; Filyushkina et al., 2019; McGregor and Nelson, 2019). One predominant theoretical framework for BG pathology in PD is the “rate model”, which posits that motor centers downstream of the BG are over-inhibited, leading to disordered movements (Albin et al., 1989; DeLong, 1990; Obeso et al., 2008; Utter and Basso, 2008; McGregor and Nelson, 2019; Vitek and Johnson, 2019). However, recent studies have revealed parkinsonian phenomena that the rate model would not predict, including the intriguing clinical observation that not all forms of movement are equally affected by PD: when movements are guided by external stimuli (e.g., gait matching with a rhythmic auditory stimulus or visually patterned flooring, kinematics are less impaired than for otherwise identical movements made in the absence of guiding stimuli (Glickstein and Stein, 1991; McIntosh et al., 1997; Ballanger et al., 2006; Daroff, 2008; McDonald et al., 2015; Distler et al., 2016). The primary question raised by this observation is whether parkinsonian BG output is similarly disrupted for these “stimulus-guided” and “internally-specified” movements. If so, we might infer that stimulus-guided movements are protected from PD via the recruitment of extra-BG pathways (Lewis et al., 2007; Hackney et al., 2015; Drucker et al., 2019; Filyushkina et al., 2019; Chen et al., 2020). However, differences in parkinsonian BG output between these forms of movements would implicate BG processing itself in this behavioral phenomenon. Given that understanding the neural basis for this clinical observation could be leveraged to improve treatment for PD, we sought to develop a suitable experimental paradigm for examining parkinsonian BG output during stimulus-guided and internally-specified movements.
We specifically sought to focus on parkinsonian BG output during movement preparation, a key motor phase in which sensory and cognitive variables are integrated (Cisek and Kalaska, 2010), and disruption of which is thought to underlie bradykinetic parkinsonian movement (Dick et al., 1984; Jahanshahi et al., 1992; Suri et al., 1998; Berardelli et al., 2001; Cutsuridis and Perantonis, 2006; Moroney et al., 2008; Wu et al., 2015; Hess and Hallett, 2017). Under normal conditions, the SNr, a BG output nucleus, is strongly engaged by the preparation of directional movements (Handel and Glimcher, 1999; Sato and Hikosaka, 2002; Lintz and Felsen, 2016). However, while numerous studies have examined parkinsonian changes in SNr activity under passive conditions or rhythmic locomotion (Hutchison et al., 1994; Wichmann et al., 1999; Galati et al., 2010; Wang et al., 2010a; Seeger-Armbruster and von Ameln-Mayerhofer, 2013; Brazhnik et al., 2014; Lobb and Jaeger, 2015; Aristieta et al., 2016; Willard et al., 2019), this approach is insufficient for differentiating how SNr activity during distinct phases of movement, including preparation, is affected by PD. To address this question, we therefore recorded SNr activity during a behavioral task in which mice with unilateral dopaminergic cell loss prepare, and subsequently initiate, SNr-engaging directional (left or right) movements that are either stimulus-guided or internally-specified (Uchida and Mainen, 2003; Thompson and Felsen, 2013; Lintz and Felsen, 2016). Crucially, by requiring that mice wait for a go signal before initiating their movement, movement preparation is temporally isolated from initiation, allowing us to dissociate the effects of PD on BG output underlying these processes.
We found that mice exhibited a directional bias ipsilateral to the hemisphere with dopaminergic cell loss, consistent with rate model predictions, and that this bias was more prominent on internally-specified than stimulus-guided trials, accordant with clinical observations. Furthermore, we found that SNr activity during movement preparation was altered in a manner consistent with the behavior, suggesting that reorganization of BG processing by dopaminergic cell loss contributes to the greater deficit in performance of internally-specified than stimulus-guided movements.
Materials and Methods
Animal subjects
All experiments were performed according to protocols approved by the University of Colorado Anschutz Medical Campus Institutional Animal Care and Use Committee. Subjects were male adult C57BL/6J mice (aged 7–14 months at the start of experiments; Jackson Labs) housed in a vivarium with a 12-hr light/dark cycle with lights on at 5:00 am. Food (Teklad Global Rodent Diet No. 2918; Harlan) was available ad libitum. Access to water was restricted prior to the behavioral session to motivate performance; however, if mice did not obtain ~1 ml of water during the behavioral session, additional water was provided for ~2–5 min following the behavioral session. All mice were weighed daily and received sufficient water during behavioral sessions to maintain >85% of pre-water restriction weight.
For behavioral analyses, only mice that completed at least 15 pre- and 15 post-surgery sessions were included (hemiparkinsonian (hemi-PD), n = 4; control, n = 4). For electrophysiological analyses, mice were included if well-isolated neurons were recorded during the task (hemi-PD, n = 5; control, n = 4). Some data from control mice were previously published using different analyses than the current study (Lintz and Felsen, 2016). For rotation assay analyses, only mice that completed at least three pre- and three-post surgery rotation assay sessions were included (hemi-PD, n = 5; control, n = 3).
Behavioral task
Mice were trained on a task requiring stimulus-guided (SG) and internally-specified (IS) movements (Fig. 1) as previously described (Lintz and Felsen, 2016). Briefly, each mouse was water-restricted and trained to interact with three ports (center: odor port; sides: reward ports (Fig. 2A) along one wall of a behavioral chamber (Island Motion). On each trial, the mouse entered the odor port, triggering the delivery of an odor; waited 488 ± 104 ms (mean ± SD) for an auditory go signal; exited the odor port; and entered one of the reward ports (Fig. 2A). Premature exit from the odor port resulted in the unavailability of reward on that trial. Odors were comprised of binary mixtures of (-)-carvone (“Odor A”) and (+)-carvone (“Odor B”). On each SG trial, one of seven odor mixtures was presented via an olfactometer (Island Motion): Odor A/Odor B = 95/5, 80/20, 60/40, 50/50, 40/60, 20/80, or 5/95. Mixtures in which Odor A > Odor B indicated reward availability only at the left port and mixtures in which Odor B > Odor A indicated reward availability only at the right port (Fig. 2B). Since we surgically targeted the left hemisphere in all mice, we refer to Odor A as the “ipsilateral odor” and Odor B as the “contralateral odor” (e.g., Fig. 2C). Similarly, we refer to the directions “left” and “right” as “ipsilateral” and “contralateral”, respectively. On trials in which Odor A = Odor B (Odor A/Odor B = 50/50), the probability of reward at the ipsilateral and contralateral ports, independently, was 0.5. Reward, consisting of 4 μl of water, was delivered by transiently opening a calibrated water valve 10–100 ms after reward port entry. Odor and water delivery were controlled, and port entries and exits were recorded, using custom software (available at https://github.com/felsenlab; adapted from C. D. Brody) written in MATLAB (MathWorks).
Experimental timeline.
Behavioral task and baseline performance. A, Port locations (top) and timing of task events (bottom). Ipsilateral (I) and contralateral (C) are defined relative to the side of brain targeted for surgery (always left). Cyan represents ipsilateral choices, magenta represents contralateral choices. B, Odor mixtures on SG (stimulus-guided) and IS (internally-specified) trials (top) and interleaving of SG and IS blocks within a session (bottom). Horizontal cyan and magenta lines indicate which port(s) were rewarded in each block: On SG trials, ipsilateral side was rewarded when Odor A > Odor B (cyan), contralateral side was rewarded when Odor B > Odor A (magenta), and either side was equally likely to be rewarded when Odor A = Odor B; on all IS trials, Odor A = Odor B and only one side was rewarded throughout the block. C, Baseline (pre-surgery) performance on SG trials for representative mouse subsequently assigned to the hemi-PD group. Gray lines show best-fit logistic functions (Materials and Methods) for each session (n = 17). Circles show average across sessions; solid line shows best-fit logistic function to all choices across sessions. D, Baseline (pre-surgery) performance on IS trials for representative mouse subsequently assigned to hemi-PD group. Gray lines link ipsilateral- and contralateral-rewarded IS blocks within the same session (n = 15). Black circles indicate medians. Mouse chose the ipsilateral port more often on ipsilateral-rewarded blocks (p = 3.05 × 10−5, 1-tailed Wilcoxon signed rank test).
Mice learned to perform SG trials (Fig. 2C) within ~48 sessions (1 session/day); detailed training stages are described in Stubblefield et al. (Stubblefield et al., 2013). Mice required an additional ~5 sessions to learn to perform interleaved blocks of SG and IS trials. On each IS trial the 50/50 mixture was presented, and reward was available only at one side throughout the block (Fig. 2D). Detailed training stages for IS trials are described in Lintz & Felsen (Stubblefield et al., 2013; Stubblefield et al., 2015; Lintz and Felsen, 2016). Mice performed 5 blocks (SG, IS, SG, IS, SG) per session (Fig. 2B); the side associated with reward switched between each IS block. Upon completing training, mice performed at least 15 sessions to establish pre-surgery baseline behavior, underwent surgery (see below), and subsequently resumed task performance (post-surgery behavior; Fig. 1).
Rotation assay
The direction of spontaneous movement was assessed before and after surgery using a standard rotation assay (Ungerstedt, 1976; Smith et al., 2012). Following intraperitoneal (i.p.) administration of d-amphetamine (2.5 mg/kg, Sigma), mice were placed in a transparent beaker with a diameter of 11.5 cm. Mice were monitored for the next 90 mins and behavior recorded using an overhead camera. Rotations were analyzed from 10 to 30 mins post i.p. injection. A rotation score was calculated by counting the total number of complete ipsilateral (left) rotations and subtracting the total number of complete contralateral (right) rotations (Fig. 3B). Repeated testing was carried out with at least 1 week between d-amphetamine injections to allow for recovery.
Validation of hemi-PD mouse model. A, Representative coronal section (0.97 mm anterior to Bregma) in hemi-PD mouse. 6-OHDA was delivered to the ipsilateral SNc. Green, Nissl; red, tyrosine hydroxylase. B, Behavior in rotation assay. Beginning 10 min. following delivery of d-amphetamine (IP, 2.5 mg/kg), net ipsilateral rotations (# ipsilateral - # contralateral) were counted for 20 min. Within mouse (n = 5 hemi-PD; n = 3 control (saline delivered to SNc instead of 6-OHDA)) change in net ipsilateral rotations between pre- and post-surgery sessions was calculated. Gray, each mouse; black, median. Hemi-PD mice exhibited more net ipsilateral rotations following surgery (p = 0.0357 1-tailed Wilcoxon rank sum test). C, Baseline activity (odor port entry to reward port exit in each trial) of SNr neurons during task performance in hemi-PD (dark gray) and control (light gray) mice. Black lines, medians. Median baseline SNr activity was higher in hemi-PD mice (p = 0.0157, 1-tailed Wilcoxon rank sum test).
SNc surgery – Unilateral 6 OHDA and saline injections
The mouse was anesthetized with isoflurane and secured in a stereotaxic device, the skull was exposed with a midline incision, and a craniotomy targeting the left SNc was performed, centered at 3.07 mm posterior from bregma, 1.250 mm lateral from the midline, and 4.35 mm deep from cortical surface (Paxinos and Franklin, 2004). Injection volume totaled 2 μL injected at target (2.43 mg 6-OHDA/mL 0.02% ascorbic acid). After suturing the incision, a topical triple antibiotic ointment (Major) mixed with 2% lidocaine hydrochloride jelly (Akorn) was applied to the scalp, the mouse was removed from the stereotaxic device, the isoflurane was turned off, and oxygen alone was delivered to the mouse to gradually alleviate anesthetic state. Mice were administered sterile isotonic saline (0.9%) for rehydration and an analgesic (Ketofen; 5 mg/kg) for pain management. Analgesic and topical antibiotic administration was repeated daily for up to 5 days, and mice were closely monitored for any signs of distress.
This procedure was identical for control mice assessed on the rotation assay but with saline injected instead of 6-OHDA. These mice did not undergo further surgery and were used solely for rotation assay testing (Fig. 3B).
SNr surgery – tetrode implantation
Details of the surgical procedure are provided in Thompson and Felsen (Thompson and Felsen, 2013). Briefly, following establishment of pre-surgery baseline behavior and (in hemi-PD mice) following unilateral 6-OHDA injections (Fig. 1), the mouse was anesthetized with isoflurane and secured in a stereotaxic device, the scalp was incised and retracted, 2 small screws were attached to the skull, and a craniotomy targeting the left SNr was performed, centered at 3.07 mm posterior from bregma and 1.25 mm lateral from the midline (Franklin and Paxinos, 2004). A VersaDrive 4 microdrive (Neuralynx), containing 4 independently adjustable tetrodes, was affixed to the skull via the screws, luting (3M), and dental acrylic (A-M Systems). A second small craniotomy was performed in order to place the ground wire in direct contact with the brain. After the acrylic hardened, a topical triple antibiotic ointment (Major) mixed with 2% lidocaine hydrochloride jelly (Akorn) was applied to the scalp, the mouse was removed from the stereotaxic device, the isoflurane was turned off, and oxygen alone was delivered to the mouse to gradually alleviate anesthetic state. Mice were administered sterile isotonic saline (0.9%) for rehydration and an analgesic (Ketofen; 5 mg/kg) for pain management. Analgesic and topical antibiotic administration was repeated daily for up to 5 days, and mice were closely monitored for any signs of distress.
Electrophysiology
Neural recordings were collected using four tetrodes, wherein each tetrode consisted of four polyimide-coated nichrome wires (Sandvik; single-wire diameter 12.5 μm) gold plated to 0.2–0.4 MΩ impedance. Electrical signals were amplified and recorded using the Digital Lynx S multichannel acquisition system (Neuralynx) in conjunction with Cheetah data acquisition software (Neuralynx). Tetrode depths were adjusted approximately 23 hrs before each recording session in order to sample an independent population of neurons across sessions. To estimate tetrode depths during each session we calculated distance traveled with respect to the rotation fraction of the screw that was affixed to the shuttle holding the tetrode. One full rotation moved the tetrode ~250 μm and tetrodes were moved ~62.5 μm between sessions. The final tetrode location was confirmed through histological assessment (see below).
Offline spike sorting and cluster quality analysis was performed using MClust software (MClust-4.3, A.D. Redish, et al.) in MATLAB. Briefly, for each tetrode, single units were isolated by manual cluster identification based on spike features derived from sampled waveforms. Identification of single units through examination of spikes in high-dimensional feature space allowed us to refine the delimitation of identified clusters by examining all possible two-dimensional combinations of selected spike features. We used standard spike features for single unit extraction: peak amplitude, energy (square root of the sum of squares of each point in the waveform, divided by the number of samples in the waveform), and the first principal component normalized by energy. Spike features were derived separately for individual leads. To assess the quality of identified clusters we calculated two standard quantitative metrics: L-ratio and isolation distance (Schmitzer-Torbert et al., 2005). Clusters with an L-ratio of less than 0.75 and isolation distance greater than 6.5 were deemed single units, which resulted in the exclusion of 7% of the identified clusters. Only clusters with few interspike intervals less than 1.5 ms were considered for further examination. Furthermore, we excluded the possibility of including data from the same neuron twice by ensuring that both the waveforms and response properties sufficiently changed across sessions. If they did not, we conservatively assumed that we were recording from the same neuron, and only included data from one session.
Immunohistochemistry
Mice were overdosed with an i.p. injection of sodium pentobarbital (100 mg/kg) and transcardially perfused with saline followed by ice-cold 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB). After perfusion, brains were submerged in 4% PFA in 0.1 M PB for 24 hrs for post-fixation and then cryoprotected for 24 hrs by immersion in 30% sucrose in 0.1 M PB. The brain was encased in the same sucrose solution, and frozen rapidly on dry ice.
Serial coronal sections (60 μm) were cut on a sliding microtome. Fluorescent Nissl (NeuroTrace, Invitrogen) was used to identify cytoarchitectural features of the SNr and verify tetrode tracks and lesion damage within or below the SNr, as previously described (Lintz and Felsen, 2016). In addition, coronal sections were stained for tyrosine hydroxylase (TH). Following repeated soaks in PBS and blocking solution, sections were exposed to primary antibody overnight (Anti-Tyrosine Hydoxylase (Rabbit) Antibody, 1:1000, Rockland). Next, sections were washed in carrier solution (2×10-min) and exposed to secondary antibody for 2 hrs (Goat anti-Rabbit IgG (H+L) Secondary Antibody, 1:500). Images were captured with a 10x objective lens, using an LSM 5 Pascal series Axioskop 2 FS MOT confocal microscope (Zeiss). For each mouse, a representative coronal section including the SNc (Paxinos & Franklin) was used to quantify dopaminergic cell loss by comparing the number of TH+ neurons ipsilateral and contralateral to the injection. Hemi-PD mice without verified >70% dopaminergic cell loss (5/10), were excluded from the group. Secondary confirmation of dopaminergic cell loss was quantified in the same manner using coronal sections containing the striatum (Fig. 3A).
Behavioral directional bias
On SG trials, directional bias was quantified using psychometric functions fitted to p = 1/ 1 + e(−a – bx), where x is the proportion of Odor A in the binary odor mixture, p is the fraction of ipsilateral choices, and a and b are the best-fit free parameters. The slope of the function is captured by b, while its directional bias was calculated as Bias = a/b + 50 (Felsen and Mainen, 2008). Positive bias values reflect an ipsilateral (leftward) shift (Fig. 4A) relative to a perfectly symmetric psychometric function, and negative values reflect a contralateral (rightward) shift (Fig. 4B). To directly compare directional bias between SG and IS trials within the same session (Fig. 6B), for SG and IS trials separately we calculated the difference between the fraction of trials for which ipsilateral was the correct choice and the fraction of actual ipsilateral choices. Sessions were included in this analysis if the mouse completed at least 25 trials of each direction (ipsilateral and contralateral) and type (SG and IS) trials.
Effect of hemi-parkinsonian state on stimulus-guided (SG) movements. A, Choices and best-fit logistic functions for SG trials in all pre- (gray, n = 30) and post- (black, n = 42) surgery sessions for a representative hemi-PD mouse. Error bars, ±SEM. Leftward arrow indicates ipsilateral directional bias post-surgery relative to pre-surgery, quantified with logistic fits (Materials and Methods). B, As in A, for a representative control mouse (tetrode drive implanted to SNr; n = 17 pre-surgery and 33 post-surgery sessions). C, Change in directional bias (post-surgery – pre-surgery) was more ipsilateral for hemi-PD than control mice (p = 0.0468, 1-tailed unpaired t-test); ipsilateral change in directional bias was significant for hemi-PD (p = 0.0478, 1-tailed t-test) but not control mice (p = 0.186, 1-tailed t-test). Horizontal bars, medians; boxes, 25th and 75th percentiles. D, Median reaction time (from go signal to reward port entry) shown separately for pre-surgery (n = 16 sessions) and post-surgery (n = 12 sessions) ipsilateral and contralateral SG trials for the same mouse as in A. Only correct trials are included; all choices on 50/50 trials are considered correct. (Error bars, ±2 SEM). E, As in D, for a representative control mouse (n = 17 pre-surgery and 33 post-surgery sessions; same representative mouse as in B). F, Change in median reaction time (post-surgery – pre-surgery) was larger for contralateral than ipsilateral correct SG trials in hemi-PD mice (p = 0.0143, 1-tailed Wilcoxon rank sum test). G, As in F, for control mice (p = 0.385, 1-tailed unpaired t-test).
Neuronal direction preference
We used an ROC-based analysis to quantify the selectivity of single neurons for movement direction (Green DM, 1966; Lintz and Felsen, 2016). This analysis calculates the ability of an ideal observer to classify whether a given firing rate was recorded in one of two conditions (i.e., preceding ipsilateral (left) or contralateral (right) movement). We defined “preference” as 2(ROCarea – 0.5), a measure ranging from −1 to 1, where −1 signifies the strongest possible preference for ipsilateral, 1 signifies the strongest possible preference for contralateral, and 0 signifies no preference (Feierstein et al., 2006; Lintz and Felsen, 2016). For example, if the firing rate of a given neuron is generally higher preceding ipsilateral than contralateral movements, that neuron is assigned a preference < 0. Statistical significance was determined with a permutation test: we recalculated the preference after randomly reassigning all firing rates to either of the two groups, repeating this procedure 500 times to obtain a distribution of values, and calculated the fraction of random values exceeding the actual value. We tested for significance at α = 0.05. Trials in which the movement time (between odor port exit and reward port entry) was > 1.5 s were excluded from all analyses. Neurons with < 25 trials of each trial type under comparison (ipsilateral SG, contralateral SG, ipsilateral IS, or contralateral IS) or with a firing rate < 2.5 spikes/s for either trial type under comparison, were excluded from all analyses.
Shift function
We used a shift function to quantify if and how two distributions differ (Rousselet et al., 2017). Briefly, using a Harrell-Davis quantile estimator (Harrell and Davis, 1982), distributions were divided into 10 equal parts by 9 “deciles.” For example, the 1st decile is the value below which 10% of the values lie while the 9th decile is the value below which 90% of values lie. The shift function compares a given decile in distribution A with its corresponding decile in distribution B. Corresponding deciles were determined to be significantly different if the confidence interval of their differences, calculated by sampling the difference between bootstrapped distributions 200 times, did not cross 0.
Activity change during epoch of interest compared to baseline
We calculated the normalized response (NR) for each neuron as NR = Ft/Fb where Ft is the median firing rate in the “test” window (either movement preparation epoch or movement initiation epoch) and Fb is the median firing rate in the “baseline” window. Since the structure of our task does not include a natural “baseline” epoch – i.e., in which the mouse is in a motionless state unaffected by task demands – our baseline window was defined as the time of odor port entry to reward port exit (i.e., the duration of the whole trial; (Lintz and Felsen, 2016). Neurons with Fb < 2.5 spikes/s were excluded from analyses. Statistical significance was determined using a pairwise t-test to compare Ft and Fb from the same trial. Neurons with NR < 1 (p < 0.05) were defined as “Decreasing” and neurons with NR > 1 (p < 0.05) were defined as “Increasing”; all other neurons were categorized as “No Δ” (Fig. 8; Table 1). Note that, by convention, a decreasing neuron that decreases more for contralateral than ipsilateral movement would be considered to have an ipsilateral direction preference (as calculated above), because firing rate is higher for ipsilateral movement (Sato and Hikosaka, 2002; Lintz and Felsen, 2016).
Direction preference and activity change from baseline for all SNr neurons recorded from hemi-PD (upper) and control (lower) mice.
Statistical Analysis
In general, MATLAB was used for all statistical analyses except for χ2 analyses, which were performed in R. Distributions were tested for normality and unless all data sets under comparison were normally distributed, non-parametric statistical tests were used. For consistency, all graphical representations of central tendency are medians, independent of whether parametric or non-parametric statistical tests were used.
Results
Effect of unilateral dopaminergic cell loss on stimulus-guided and internally-specified movements
To examine how parkinsonian conditions affect stimulus-guided (SG) and internally-specified (IS) movements and their underlying BG output, we first trained mice on a behavioral task designed to elicit these forms of movements (Figs. 1, 2A,B) (Lintz and Felsen, 2016). Briefly, on each trial of the task, the mouse was presented with a binary odor mixture at a central port, waited for an auditory go signal, and moved to the ipsilateral (left) or contralateral (right) reward port for a water reward (Fig. 2A). Each daily experimental session consisted of interleaved blocks of SG and IS trials (Fig. 2B; Materials and Methods)(Lintz and Felsen, 2016). On SG trials, the dominant component of the odor mixture – which varied by trial – determined the side at which reward was delivered, while on IS trials, a balanced mixture of the two odors preceded delivery of reward at the same side throughout the block (Fig. 2B; Materials and Methods). Consistent with previous results (Lintz and Felsen, 2016; Lintz et al., 2019), the direction of movement on SG trials was selected based on the stimulus (Fig. 2C) while the direction of movement on IS trials was selected based on recent trial history (Fig. 2D).
Upon achieving proficient performance on the task (Fig. 2C,D; Materials and Methods), mice received 6-OHDA injections to one SNc to unilaterally ablate dopaminergic neurons (Ungerstedt, 1976; Chang et al., 2006; Israel and Bergman, 2008; Avila et al., 2010; Smith and Heuer, 2011; Brazhnik et al., 2012; Seeger-Armbruster and von Ameln-Mayerhofer, 2013; Brazhnik et al., 2014) and were implanted with a chronic tetrode drive targeting the ipsilateral SNr to record BG output (Fig. 1). Only mice with > 70% histologically-confirmed dopaminergic cell loss (examined following all behavioral and recording experiments; Fig. 1, Fig. 3A) were included in the “hemi-PD” group for all subsequent analyses (5/10 mice; Materials and Methods). Confirming the validity of our hemi-PD model, hemi-PD mice exhibited a greater ipsilateral bias than control mice (saline delivered to SNc) on a standard rotation assay (p = 0.0357, 1-tailed Wilcoxon rank sum test; Fig. 3B; Materials and Methods), and the mean firing rate of SNr neurons was higher (p = 0.0157, 1-tailed Wilcoxon rank sum test; Fig. 3C), consistent with rate-model predictions and previous reports (Sanderson et al., 1986; Hutchison et al., 1994; Wichmann et al., 1999; Chang et al., 2006; Brazhnik et al., 2014; Lobb and Jaeger, 2015).
We next examined whether unilateral dopaminergic cell loss affected the SG and IS movements required by our task (Fig. 1). Given the rate-model prediction that SNr-recipient nuclei for contralateral movement are over-inhibited (Albin et al., 1989; DeLong, 1990; Obeso et al., 2008; Utter and Basso, 2008; McGregor and Nelson, 2019; Vitek and Johnson, 2019), and supported by our findings of greater ipsilateral bias on the rotation assay (Fig. 3B) and an increase in mean SNr activity (Fig. 3C), we expected hemi-PD mice to exhibit an ipsilateral bias for both SG and IS movements. To quantify the extent of directional bias on SG trials, for each mouse we estimated the pre-surgery and post-surgery bias based on a logistic function fit separately to the choices from all sessions before and all sessions after surgery, respectively (Fig. 4A,B; Materials and Methods). In hemi-PD mice, we found that choices after surgery were biased ipsilaterally compared to before surgery (p = 0.0478, 1-tailed unpaired t-test; Fig. 4A,C), while no post-surgery change in bias was observed in control mice (p = 0.186, 1-tailed unpaired t-test; post-surgery change in ipsilateral bias was greater for hemi-PD than control mice, p= 0.0468, 1-tailed unpaired t-test; Fig. 4B,C). Consistent with their ipsilateral directional bias, hemi-PD mice exhibited a larger increase in reaction time (defined as the time from go signal to reward port entry) post-surgery for contralateral than ipsilateral movements (p = 0.0143, 1-tailed Wilcoxon rank sum test; Fig. 4D,F). While control mice also exhibited longer reaction times post-surgery, as expected given the additional weight of the chronic recording drive, the increase was not greater for contralateral than ipsilateral movements (p = 0.385, 1-tailed unpaired t-test; Fig. 4E,G).
We observed similar effects of unilateral dopaminergic cell loss on IS movements (Fig. 5). For these trials, we quantified bias by separately examining the fraction of ipsilateral choices during blocks in which reward was delivered at the ipsilateral port (“ipsilateral blocks”) and at the contralateral port (“contralateral blocks”). On ipsilateral blocks, both groups of mice nearly always chose the ipsilateral port (median fraction ipsilateral choices: hemi-PD pre-surgery = 0.88, post-surgery = 0.85; control pre-surgery = 0.89, post-surgery = 0.87; Fig. 5A,B). On contralateral blocks, however, hemi-PD mice were more likely to choose the ipsilateral port after than before surgery (median pre-surgery = 0.12, post-surgery = 0.31; p = 0.0115, 1-tailed paired t-test; Fig. 5A,C), while control mice did not exhibit this change (median pre-surgery = 0.12, post-surgery = 0.18; p = 0.100, 1-tailed paired t-test; post-surgery change in ipsilateral bias was greater for hemi-PD than control mice, p = 0.0298, 1-tailed unpaired t-test; Fig. 5B,C). We also found the same effect on reaction times on IS trials as we did on SG trials: hemi-PD (but not control) mice exhibited a greater post-surgery increase in reaction time on contralateral than ipsilateral trials (hemi-PD: p = 0.0143, 1-tailed Wilcoxon rank sum test; control: p = 0.434, 1-tailed unpaired t-test; Fig. 5D-G). Thus, on both SG and IS trials, unilateral dopaminergic cell loss resulted in fewer and slower contralateral movements, consistent with rate-model predictions.
Effect of hemi-parkinsonian state on internally-specified (IS) movements. A. Separately for IS trials in which the ipsilateral and contralateral side was rewarded, median fraction ipsilateral choices for pre-surgery (n = 65 sessions) and post-surgery (n = 22 sessions) sessions are shown for a representative hemi-PD mouse. Fraction of ipsilateral choices increased post-surgery relative to pre-surgery on trials in which the contralateral side was rewarded, but not on trials in which the ipsilateral side was rewarded (Error bars, ±2 SEM). B, As in A, for a representative control mouse (tetrode drive implanted; n = 14 pre-surgery and 13 post-surgery sessions) C. Change in fraction ipsilateral choices on trials in which contralateral side was rewarded (post-surgery – pre-surgery) was more ipsilateral for hemi-PD than control mice (p = 0.0298, 1-tailed unpaired t-test); ipsilateral change was significant for hemi-PD (p = 0.0115, 1-tailed t-test) but not control mice (p = 0.1004, 1-tailed t-test). Horizontal bars, medians; boxes, 25th and 75th percentiles. D, Median reaction time (from go signal to reward port entry) shown separately for pre-surgery (n = 65 sessions) and post-surgery (n = 22 sessions) ipsilateral and contralateral IS trials for the same mouse as in A). Only correct trials are included. (Error bars, ±2 SEM). E, As in D, for a representative control mouse (n = 14 pre-surgery and 13 post-surgery sessions, same representative mouse as in B). F, Change in median reaction time (post-surgery – pre-surgery) was larger for contralateral than ipsilateral IS trials in hemi-PD mice (p = 0.0143, 1-tailed Wilcoxon rank sum test). G, As in F, for control mice (p = 0.434, 1-tailed unpaired t-test).
We next asked whether this effect was greater in magnitude, indicative of greater impairment, on SG or IS trials. Given that PD patients are less impaired when performing movements guided by external stimuli (Glickstein and Stein, 1991; Ballanger et al., 2006; Daroff, 2008; McDonald et al., 2015; Distler et al., 2016; Tekriwal et al., 2018), we expected unilateral dopaminergic cell loss to have a greater impact on IS trials. To examine this possibility, we directly compared, within each session, behavior on SG and IS trials. We first compared the percent correct, as a measure of overall performance on these two trial types. We found that, before surgery, hemi-PD mice performed better (higher percent correct) on IS trials, consistent with the observation that the correct choice remained the same throughout the block (Fig. 6A). However, post-surgery, mice exhibited a significant shift toward better relative performance on SG trials (p = 6.52 × 10−15, 1-tailed Wilcoxon rank sum test; Fig. 6A). We next compared directional bias on SG and IS trials by calculating the difference between the fraction of trials for which ipsilateral was the correct choice and the fraction of actual ipsilateral choices, separately for SG and IS blocks within each session. We found that, pre-surgery, hemi-PD mice were slightly more ipsilaterally biased on IS than SG trials, but post-surgery they exhibited a greater ipsilateral bias on IS than SG trials (p = 0.0300, 1-tailed unpaired t-test; Fig. 6B). Finally, we compared median reaction times between SG and IS trials within each session, separately for ipsilateral and contralateral movements. To control for the potential effects of discrimination difficulty on reaction time, we only included “easy discrimination” SG trials (95:5, 80:20, 20:80, 5:95)(Lintz and Felsen, 2016) in this analysis. We found that, pre-surgery, reaction times in both directions tended to be somewhat shorter on IS than SG trials, consistent with previous findings (Lintz and Felsen, 2016) (ppre_ipsi = 1.25 x 10−12; ppre_contra = 4.65 x 10−12; 1-tailed Wilcoxon sign rank test; Fig. 6C). Post-surgery, reaction times for ipsilateral movements exhibited this same pattern, but there was no difference in reaction times between contralateral SG and IS movements (ppost_ipsi = 6.10 x 10−5; ppost_contra = 0.0901, 1-tailed Wilcoxon sign rank test). Together, these analyses demonstrate that IS trials are relatively more affected than SG trials by unilateral dopaminergic cell loss.
Direct comparison between stimulus-guided (SG) and internally-specified (IS) behavior in hemi-PD mice. A, Within-session comparison of % correct between SG and IS trials for pre- and post-surgery sessions for hemi-PD mice (n = 4). All 50/50 SG trials were considered correct. Pre-surgery performance was better on IS trials; post-surgery (hemi-parkinsonian) performance was better on SG trials (p = 6.52 × 10−15, 1-tailed Wilcoxon rank sum test). Horizontal bars, medians; boxes, 25th and 75th percentiles. (Controls, not pictured, p = 0.405, 1-tailed Wilcoxon rank sum test, n = 153 pre-surgery sessions, n = 48 post-surgery sessions). B, Within-session comparison of fraction of ipsilateral choices between SG and IS trials for all hemi-parkinsonian sessions, see Materials and Methods for calculation details. Hemi-parkinsonian behavior was more ipsilaterally biased on IS than SG trials as compared to pre-surgery (p = 0.0300, 1-tailed unpaired t-test) (Controls, not pictured, p = 0.707, 1-tailed unpaired t-test, n = 153 pre-surgery sessions, n = 48 post-surgery sessions). C, Within-session comparison of reaction time between SG and IS trials for all hemi-parkinsonian sessions. In pre-surgery sessions, reaction times were faster on IS than SG trials for both directions. In hemi-parkinsonian sessions, reaction times were faster on IS than SG ipsilateral trials, but did not differ between IS and SG contralateral trials. To control for multiple comparisons, we first ran a Kruskal-Wallis as an ANOVA measure. If this test rejected the null that all groups are from the same distribution, we then investigated specific comparisons based on our planned analyses (p = 9.83 x 10−7, Kruskal-Wallis test; ppre_ipsi = 1.25 x 10−12, 1-tailed Wilcoxon sign rank test; ppost_ipsi = 6.10 x 10−5, 1-tailed Wilcoxon sign rank test; ppre_contra = 4.65 x 10−12, 1-tailed Wilcoxon sign rank test; ppost_contra = 0.0901, 1-tailed Wilcoxon sign rank test). (Controls, not pictured, p = 5.11 x 10−6, Kruskal-Wallis test; ppre_ipsi = 9.95 x 10−10, 1-tailed Wilcoxon sign rank test; ppost_ipsi = 7.27 x 10−7, 1-tailed Wilcoxon sign rank test; ppre_contra = 7.12 x 10 −23, 1-tailed Wilcoxon sign rank test; ppost_contra = 2.41 x 10−4, 1-tailed Wilcoxon sign rank test).
Effect of unilateral dopaminergic cell loss on basal ganglia output
We sought to determine whether behavioral differences between SG and IS trials could be accounted for by activity in the SNr, which is known to play a role in the movements elicited by this task (Lintz and Felsen, 2016). In the same mice described in the above behavioral results, we used chronically implanted tetrodes to record from 183 SNr neurons in hemi-PD mice (n = 5) and 285 neurons in control mice (n = 4) that met our analysis criteria (Materials and Methods). We focused on activity underlying movement preparation, which has been implicated in parkinsonian motor deficits (Dick et al., 1984; Jahanshahi et al., 1992; Suri et al., 1998; Berardelli et al., 2001; Cutsuridis and Perantonis, 2006; Moroney et al., 2008; Wu et al., 2015; Hess and Hallett, 2017). Consistent with previous results (Lintz and Felsen, 2016), we observed that SNr activity recorded from hemi-PD mice was often modulated during the “movement preparation epoch” (from 100 ms after the odor valve was opened until the go signal) and depended on movement direction (Fig. 7A). We first asked whether firing rate during this epoch differed between hemi-PD and control mice, similar to the difference we and others have observed in baseline activity (Fig. 3C)(Hutchison et al., 1994; Wichmann et al., 1999; Galati et al., 2010; Wang et al., 2010a; Seeger-Armbruster and von Ameln-Mayerhofer, 2013; Brazhnik et al., 2014; Lobb and Jaeger, 2015; Aristieta et al., 2016; Willard et al., 2019). The rate model would predict elevated SNr activity in hemi-PD mice, consistent with the ipsilateral bias that we observed given that SNr activity inhibits downstream motor centers that primarily mediate contralateral movement. However, we found no overall difference between groups for movement in either direction (pipsi= 0.0520; pcontra= 0.0954, 1-tailed Wilcoxon rank sum tests; Fig. 7B,C). The ipsilateral bias exhibited by hemi-PD mice on SG and IS trials therefore cannot be explained simply by an absolute increase in SNr activity during movement preparation.
SNr activity during movement preparation in behaving hemi-PD and control mice. A, Rasters (top) and peri-event time histograms (bottom) for an example neuron from a hemi-PD mouse aligned to odor valve open and segregated by choice. Histograms are smoothed with a Gaussian filter; shading, ±SEM. B,C, Mean firing rate during movement preparation epoch (between odor valve open and go signal) did not differ between populations of neurons in hemi-PD and control mice on ipsilateral (B) or contralateral (C) trials (pipsi= 0.0520; pcontra= 0.0954, 1-tailed Wilcoxon rank sum tests; n = 183 neurons in 5 hemi-PD mice; 285 neurons in 4 control mice). Horizontal bars, medians. D, Distribution of direction preferences for population of neurons in hemi-PD mice (top) had a smaller range than in control mice (bottom) (p = 7.31 × 10−4, 2-sample Kolmogorov-Smirnov test), and more neurons exhibited a significant preference in control than in hemi-PD mice (p = 2.20 × 10−16, χ2-test). Arrowhead, example neuron shown in A.
We next examined whether unilateral dopaminergic cell loss affected the relative activity of individual neurons between ipsilateral and contralateral movements (Fig. 7A), which could also potentially account for the ipsilateral behavioral bias. We therefore calculated the “direction preference” of each neuron, which quantifies the difference in firing rate during a specified epoch between ipsilateral and contralateral movements, and ranges from −1 (higher firing rates for ipsilateral movements) to 1 (higher firing rates for contralateral movements), where 0 represents no preference (Materials and Methods)(Green DM, 1966; Feierstein et al., 2006; Lintz and Felsen, 2016). We found that the populations of neurons recorded in hemi-PD and control mice each exhibited a range of preferences, with some neurons preferring ipsilateral movement (Fig. 7D, cyan; p < 0.05, permutation test; Materials and Methods) and some preferring contralateral movement (Fig. 7D, magenta; p < 0.05, permutation test). However, the distributions of preferences exhibited by neurons recorded in hemi-PD and control mice differ (p = 7.31 × 10−4, 2-sample Kolmogorov-Smirnov test) in two key respects.
First, the population of SNr neurons in hemi-PD mice exhibited weaker direction preference than the population in control mice. We quantified this difference with several complementary analyses. A smaller proportion of neurons in hemi-PD mice (52/183, 28%; Table 1) than control mice (199/285, 70%; Table 1) exhibited a significant direction preference (p = 2.20 × 10−16, χ2-test = 147, df = 1). When the entire population in hemi-PD and control mice is considered (i.e., including neurons with and without a significant direction preference), the strength of the preference, independent of sign, was smaller in hemi-PD mice (median = 0.0815) than control mice (median = 0.179; p = 1.87× 10−12, 1-tailed Wilcoxon rank sum tests). Finally, we used a shift function analysis to identify the deciles in which the distributions differed (Rousselet et al., 2017) (Materials and Methods). This analysis revealed that preferences were closer to 0 in hemi-PD than control mice in the 3 deciles representing the most ipsilateral preferences and in the 3 deciles representing the most contralateral preferences, consistent with weaker direction preference in hemi-PD mice. Together, these analyses indicate that unilateral dopaminergic cell loss results in a fundamental disruption of the representation of movement direction that is normally observed in the SNr (Handel and Glimcher, 1999; Berardelli et al., 2001; Lintz and Felsen, 2016).
Second, the population of SNr neurons in hemi-PD mice, but not control mice, exhibited a slight bias towards contralateral preferences. While the entire distribution of preferences in hemi-PD mice was not contralaterally skewed (median = 0.136; p = 0.368, 1-tailed Wilcoxon sign-rank test), of the neurons that exhibited a significant preference (Fig. 7D, magenta and cyan, Table 1), more were contralateral-preferring (32/52, magenta; Table 1) than ipsilateral-preferring (20/52, cyan; p = 0.0155, χ2-test = 4.65 df = 1; Table 1). In contrast, in control mice we found roughly equal proportions of contralateral-preferring (105/199, magenta; Table 1) and ipsilateral-preferring (94/199, cyan; Table 1) neurons (p = 0.158, χ2-test = 1.01, df = 1; Fig. 7D, Table 1). This analysis indicates that, in the population of SNr neurons in which direction preference is spared, unilateral dopaminergic cell loss results in more neurons exhibiting higher activity for contralateral than ipsilateral movements, consistent with the ipsilateral behavioral bias that we observed (Figs. 4,5, Table 1).
To gain insight into the reorganization of BG output induced by unilateral dopaminergic cell loss, we next examined these systematic differences in preferences between hemi-PD and control mice within functional classes of SNr neurons (Fig. 8, Table 1). Separate subpopulations of SNr neurons are known to exhibit increases or decreases in activity as movements are prepared and initiated (Handel and Glimcher, 1999; Sato and Hikosaka, 2002; Lintz and Felsen, 2016); these subpopulations presumably play different functional roles. We therefore categorized neurons into one of three classes based on whether their activity during movement preparation increased, decreased, or did not change, relative to baseline (Table 1; Materials and Methods). We first noticed a clear difference in the proportion of neurons in each class between hemi-PD and control groups (p = 1.32 × 10−11, χ2-test = 50.108, df = 2; n = 183 neurons, 5 hemi-PD mice; n = 285 neurons, 4 control mice; Fig. 8A, Table 1). Specifically, we found that a higher proportion of neurons recorded in hemi-PD mice (54.1%) than in control mice (40.0%) exhibited decreased activity, and a corresponding lower proportion of neurons recorded in hemi-PD mice (19.7%) than in control mice (44.9%) exhibited increased activity (Fig. 8A, Table 1; Materials and Methods). In addition, our finding that fewer neurons in hemi-PD than control mice exhibited a significant direction preference held true across all 3 subpopulations of neurons (Fig. 8A, Table 1). Finally, we found that the contralateral bias in the hemi-PD group among neurons exhibiting a direction preference was largely due to those that exhibited decreased activity during movement preparation (contralateral:ipsilateral ratio = 25:7, p = 1.07 × 10−5, χ2-test = 18.1, df = 1; Fig. 8A, Table 1); this subpopulation in the control group did not show this effect (contralateral:ipsilateral ratio = 36:42; p = 0.788, χ2-test = 0.641, df = 1; Fig. 8A, Table 1). Thus, unilateral dopaminergic cell loss resulted in a larger proportion of SNr neurons that release downstream motor centers from inhibition, with a greater effect on ipsilateral than contralateral movements, accounting for the relationship between SNr activity and the ipsilateral behavioral bias.
Functional classes of SNr neurons during movement preparation and initiation in behaving hemi-PD and control mice. A, Fraction of neurons with a given direction preference segregated by whether their average activity during the movement preparation epoch (beginning 100 ms after the odor valve opens and ending with the go signal) significantly increased (Inc.), decreased (Dec.) or did not change (No Δ) relative to average baseline firing rate, for hemi-PD (top) and control (bottom) mice. Proportion of neurons exhibiting an increase, a decrease, and no change differs between hemi-PD and control mice (p = 1.32 × 10−11, χ2-test = 50.108, df = 2). Similarly, proportion of ipsilateral, contralateral, and no direction preference neurons differed between hemi-PD and control mice (p = 2.2 × 10−16, χ2-test = 149.58, df = 2). B, Fraction of neurons with a given direction preference segregated by whether their activity during the movement initiation epoch (beginning with the go signal and ending 100 ms after odor poke out) increased, decreased or did not change relative to pre-surgery, for hemi-PD (top) and control (bottom) mice. Proportion of neurons exhibiting an increase, a decrease, and no change did not differ between hemi-PD and control mice (p = 0.459, χ2-test = 1.5586, df = 2; n = 183 neurons in 5 hemi-PD mice; 285 neurons in 4 control mice); proportion of ipsilateral, contralateral, and no direction preference neurons differed between hemi-PD and control mice (p = 4.413 × 10−5, χ2-test = 20.049, df = 2).
Thus far we have focused on SNr activity during the movement preparation epoch. As our task is designed to separate movement preparation from initiation, we were able to extend our analyses to movement initiation. We expected to observe similar changes during movement initiation, consistent with previous studies (Kravitz et al., 2010; Wang et al., 2010b; Abedi et al., 2013; Freeze et al., 2013). We therefore repeated our firing rate, direction preference and functional class analyses for the period from the go signal until 100 ms after odor port exit, when the movement is initiated. As with the movement preparation epoch, we found no overall difference in firing rate during movement initiation between the hemi-PD and control groups for movement in either direction (pipsi= 0.193; pcontra= 0.103, 1-tailed Wilcoxon rank sum tests). Next, as with movement preparation, we found that a smaller proportion of neurons in hemi-PD than control mice exhibited a direction preference during movement initiation (p = 4.41 × 10−5, χ2-test = 20.0, df = 2; Fig. 8B), but our shift function analysis comparing the distributions of direction preferences in hemi-PD and control mice revealed only one differing decile (8th decile). Consistent with this result, we found no difference between hemi-PD and control mice in the proportion of neurons that increased or decreased activity during movement initiation compared to baseline (p = 0.459, χ2-test = 1.56, df = 2; n = 183 neurons, 5 hemi-PD mice; n = 285 neurons, 4 control mice; Fig. 8B, Table 1). Thus, dopaminergic cell loss appears to affect BG output more during movement preparation than during movement initiation in the context of our behavioral task.
Finally, we examined whether the changes in SNr activity during movement preparation associated with unilateral dopaminergic cell loss were consistent with the stronger ipsilateral behavioral bias on IS compared to SG trials (Fig. 6). For example, the representative neuron shown in Fig. 9A appears to exhibit a contralateral preference, consistent with an ipsilateral behavioral bias, on IS but not SG trials. To examine this phenomenon across the population, we compared direction preference during the movement preparation epoch between IS and SG trials within the same session (Fig. 9B). In the hemi-PD group, we found that preference was significantly greater (i.e. more contralateral) on IS than SG trials (p = 0.00390, 1-tailed Wilcoxon sign rank test, n = 158 neurons; Fig. 9B). We also observed a significant difference in control mice (p = 0.0180, 2-tailed Wilcoxon sign rank test; n = 285 neurons), but in the opposite direction (i.e. more ipsilateral). Together, these analyses of SNr activity show that changes in BG output in hemi-PD mice are consistent with their overall ipsilateral behavioral bias (Figs. 4,5), as well as their stronger bias on IS than SG trials (Fig. 6).
SNr activity on stimulus-guided (SG) and internally-specified (IS) trials in hemi-PD mice. A, Rasters (top) and peri-event histograms (bottom), on stimulus-guided (SG, left) and internally-specified (IS, right) trials, for an example neuron from a hemi-PD mouse aligned to odor valve open and segregated by choice. Histograms are smoothed with a Gaussian filter; shading, ±SEM. B, Direction preference on SG and IS trials for population of SNr neurons in hemi-PD mice. Each neuron is represented by a pair of connected gray symbols; only neurons with significant preference (p < 0.05) on SG and/or IS trials are shown. Preference was more contralateral on IS than SG trials (p = 0.0132, 1-tailed Wilcoxon sign rank test, n = 79). Black symbols, medians.
Discussion
In this study, we primarily sought to determine whether BG output could explain the greater impairment in IS than SG movements under parkinsonian conditions. By recording SNr activity in hemi-PD mice performing SG and IS movements, we found that unilateral dopaminergic cell loss alters the relationship between SNr activity and movement direction as movements are prepared (Figs. 7D, 8, Table 1), consistent with the ipsilateral bias in behavior (Figs 4, 5). While we did not observe an absolute increase in preparation-related SNr activity in hemi-PD mice (Fig. 7B,C), as predicted by classical models of parkinsonian BG activity (DeLong, 1990; McGregor and Nelson, 2019), our findings are consistent with a rate-model framework in which greater SNr output inhibits downstream motor nuclei mediating contralateral movements (DeLong, 1990; McGregor and Nelson, 2019; Vitek and Johnson, 2019). In contrast, neural activity during movement initiation was little changed between hemi-PD and control groups (Fig. 8B), suggesting that parkinsonian conditions affect BG output subserving movement preparation more than initiation, consistent with studies in PD patients (Jahanshahi et al., 1992; Beiser and Houk, 1998; Suri et al., 1998; Cutsuridis and Perantonis, 2006; Moroney et al., 2008; Wu et al., 2015).
While unilateral dopaminergic cell loss resulted in an ipsilateral bias on both SG and IS trials (Figs. 4 and 5), we found that the effect was larger on IS trials (Fig. 6). This difference was reflected in the activity of SNr neurons, which exhibited a stronger contralateral preference (i.e., higher activity on contralateral than ipsilateral trials) on IS than SG trials (Fig. 9). This finding is consistent with our previous work showing that SNr activity is more sensitive to movement direction under IS than SG conditions (Lintz and Felsen, 2016), which suggests that the cognitive processes associated with IS movements are more BG-dependent than those associated with SG movements. One key factor that may explain the difference in BG-dependence between IS and SG trials is the influence of prior choices and outcomes, which are crucial for determining the correct direction of movement on IS, but not SG, trials. The SNr receives direct excitatory input from the pedunculopontine tegmental nucleus (Scarnati et al., 1984; Beninato and Spencer, 1987), which encodes prior choices and outcomes (Thompson and Felsen, 2013); SNr activity itself reflects prior choices (Lintz and Felsen, 2016), and in general the BG are thought to bias activity in downstream motor centers towards movements associated with larger rewards (Hikosaka et al., 2000; Sato and Hikosaka, 2002; Watanabe et al., 2003; Kawagoe et al., 2004; Hikosaka et al., 2006). The disruption of BG signaling by unilateral dopaminergic cell loss may therefore affect the (adaptive) influence of priors on movement selection (Perugini et al., 2016), resulting in a greater deficit on IS than SG trials. The importance of the BG in representing priors is also consistent with our finding that SNr activity related to movement preparation, which is influenced by priors, was more affected than activity related to movement initiation, which is not.
In addition to the shift in the population of neurons in hemi-PD mice towards contralateral preference consistent with the ipsilateral behavioral bias, a large proportion of neurons exhibited no relationship between activity and movement direction (gray bars in Figs. 7D, 8A, Table 1). Further, we found that more neurons in hemi-PD mice exhibited a decrease than an increase in activity during movement preparation compared to baseline activity, which was not the case in control mice (Fig. 8A, Table 1), perhaps due to the increased baseline activity in hemi-PD mice (Fig. 3C). While these finding do not directly relate to our primary behavioral readout of directional bias, they suggest a profound reorganization of BG output under parkinsonian conditions. How this reorganization could contribute to other features of parkinsonian behavior can be examined in future studies.
While we have demonstrated a clear link between the electrophysiological and behavioral effects of unilateral dopaminergic cell loss, it is worth considering potential caveats in interpreting our results. First, the hemi-PD mice included in our study all exhibited an ipsilateral behavioral bias. Although our electrophysiological results are consistent with this bias, we would be able to draw stronger conclusions about the relationship between BG output and behavior with a set of mice exhibiting a wider range of behavioral biases. Indeed, one mouse that was excluded from our hemi-PD group due to insufficient dopaminergic cell loss (< 70%; Fig. 3A) exhibited a significant contralateral bias on both SG and IS trials (SGΔ Directional bias = −12.9, see Fig. 2C for comparisons; ISΔ Fraction ipsi. choices = −0.0792, see Fig. 3C for comparisons). When we analyzed the SNr recordings from this mouse, we found that the neurons with a significant direction preference (74/83 neurons) largely exhibited an ipsilateral preference during the movement preparation epoch (ipsilateral:contralateral ratio = 59:15, p = 7.74 × 10−13, χ2-test = 50.0, df = 1, see Fig. 8A, Table 1 for comparisons). While we can only speculate about the potential compensatory adaptations that emerge in the BG with relatively moderate dopaminergic cell loss, the fact that the direction of behavioral bias remains consistent with neural direction preference further supports our finding that BG output mediates the effect of parkinsonian conditions on behavior.
Second, olfactory deficits are a known hallmark of PD (Hawkes and Shephard, 1993; Hawkes, 1995; Tarakad and Jankovic, 2017; Tekriwal et al., 2017). Given our use of an olfactory task, is it possible that the behavioral effects under parkinsonian conditions are due to sensory and not motor deficits. We suggest that this is unlikely for two reasons: 1) on IS trials the olfactory cue was not informative about reward location, 2) and on SG trials a deficit in olfactory discrimination would be reflected in a flattening, rather than a shift, in the psychometric function, which we did not observe.
Finally, while the hemi-PD model provides a powerful approach for examining the neural basis of parkinsonian movement (Ungerstedt, 1976; Avila et al., 2010; Galati et al., 2010; Brazhnik et al., 2012; Brazhnik et al., 2014; Dorval and Grill, 2014), PD typically presents with bilateral dopaminergic cell loss. It is possible that dopaminergic input from the spared SNc may compensate for the unilateral insult, and other differences between the model and the clinical condition must be considered. However, our comparison between ipsilateral and contralateral movements, at the behavioral and neurophysiological levels, was well suited to the hemi-PD model, and we suggest that our results can cautiously inform the reorganization of BG output that occurs in PD.
In conclusion, we found that the behavioral effects of unilateral dopaminergic cell loss, including differences between stimulus-guided and internally-specified movements, can be accounted for by changes in SNr activity during movement preparation. While our results could not be explained by the simplest prediction of the classical rate model that BG output is tonically elevated by dopaminergic cell loss, they were consistent with a model in which output during movement preparation determines movement direction. Future studies can examine how the BG interacts with other motor systems to differentially mediate stimulus-guided and internally-specified movements under parkinsonian conditions.
The authors declare no competing financial interests.
Acknowledgments
This work was supported by the National Institutes of Health (R01NS079518) and the Boettcher Foundation’s Webb-Waring Biomedical Research Award. Technical support was provided by the Optogenetics and Neural Engineering Core at the University of Colorado School of Medicine, funded in part by the National Institutes of Health (P30NS048154). We thank Quang Dang and Ben Peterson for data collection, and members of the Felsen lab for comments on the manuscript.